K
In your code, the return https://www.php.net/manual/en/function.preg-last-error.php It was 6, which corresponds to the error https://www.php.net/manual/en/pcre.constants.php#constant.preg-jit-stacklimit-error . Basically, this error refers to http://www.pcre.org/original/doc/html/pcrejit.html , which is a feature that makes multiple optimizations in a regex. But these optimizations are not for free: they need extra memory for their internal structures, and this error occurs when regex ends up using more memory than the availableby default, http://www.pcre.org/original/doc/html/pcrejit.html#SEC8 ).The JIT PCRE https://www.php.net/manual/en/pcre.configuration.php#ini.pcre.jit , and researching found several links that suggest https://stackoverflow.com/a/34850443 No php.ini (setting) pcre.jit = 0) or directly in the code, calling ini_set('pcre.jit', false). But I think the best option is to optimize regex so that it is more efficient, thus consuming less resources and avoiding stack JIT'sFor starters, your regex makes use of parentheses, which creates https://www.regular-expressions.info/brackets.html , which consume extra memory, since it is a structure more to be generated and maintained (this also consumes more memory when searching for the matches, because the groups must be kept separately). But since you are not using the groups (and simply replacing everything with another string), you can remove them.Another point is the https://www.regular-expressions.info/alternation.html (.|\n). The point is https://www.regular-expressions.info/dot.html , and maybe that's why you put the |\n later. The problem is that this is also inefficient, because for each character the test is done (if it corresponds to one or another).Fortunately, in various languages and engines, there is an option that makes the point also match line breaks, and the difference is brutal. See https://regex101.com/r/T73TOv/1/ , and notice the amount of steps executed (over 11,000). Already using the flag s (which makes the point match line breaks), the https://regex101.com/r/T73TOv/2/ and needs little more than 2600 steps (about 4 times less steps).You can still improve a little more. By default, the quantifiers - as the * - https://www.rexegg.com/regex-quantifiers.html#greedytrap and try to get as much characters as possible. As we are using the point, which takes any character, it tries to go to the end of the string, and then goes back to find something that satisfies the rest of the regex (process known as backtracking, which also consumes more resources, since engine need to keep the states in memory until all possibilities are exhausted).To avoid this, you can use .*?, which makes the quantifier https://www.rexegg.com/regex-quantifiers.html#lazy_solution . Thus, it takes as little characters as possible, and goes advancing gradually in the string (instead of picking up everything and going back). This drastically decreases the amount of steps performed (see that in this case, https://regex101.com/r/T73TOv/3/ ).In short, your regex could stay like this:$pattern = "/{$tA}.*?{$tB}/s";
I removed the parentheses, since I'm not using the groups for anything. I used it. https://www.php.net/manual/en/reference.pcre.pattern.modifiers.php so that the point also matches line breaks, and used the lazy quantifier *? to decrease backtracking.On my machine I did not have the same mistakes as you, but testing on the IdeOne. with we can see that https://ideone.com/5r4NNB , and changing to the regex above, https://ideone.com/POLxah .In short, his suspicion of the amount of characters was not completely unfounded. After all, the greater the string, the more backtracking will be necessary for regex to find the matches. And by improving regex, we can reduce backtracking, and consequently, the resources used to run it.Do not use regexBut perhaps the main problem is the fact that you are processing an HTML file with regex - since https://stackoverflow.com/a/1732454 . Although regex can work, in many cases it is better to use a specific API. In case, you could use https://www.php.net/manual/en/class.domdocument.php :$dom = new DOMDocument;
$dom->loadHtml($html); // $html é uma string contendo todo o HTML
$xpath = new DOMXPath($dom);
for ($i=0; $i < 11; $i++) {
foreach ($xpath->query('//comment()') as $comment) { // procura comentários
if ($comment->nodeValue == "a$i") { // comentário inicial
$parent = $comment->parentNode;
$remover = [];
// percorrer os nós irmãos até encontrar o comentário que fecha
$node = $comment->nextSibling;
while (true) {
$remover[] = $node;
$node = $node->nextSibling;
if ($node->nodeType == XML_COMMENT_NODE && $node->nodeValue == "$i") {
$remover[] = $node;
break;
}
}
foreach($remover as $n) { // remover
$parent->removeChild($n);
}
// trocar o comentário "aX" por "a"
$parent->replaceChild($dom->createComment('a'), $comment);
}
}
}
// imprimir o HTML final (com as tags removidas)
echo $dom->saveHTML();
So we can go through the HTML looking for comments. And when the comment matches what you want to find, just go through the rest of us until you find the comment that closes the passage to be censored, and remove all (in the end, I still replace the comment text, just as regex was doing).
